Published · Updated
Drafting from the matter, not from a prompt
A drafting tool that starts from a blank box and a text prompt gives you text; it doesn't give you a draft grounded in your case. AI drafting works differently: context flows in first — strategy signals from matter intelligence, evidence from the vault, the attorney's instructions — and the document comes out bound to those facts, using jurisdiction-specific rules.
The output arrives ready for review: citations, a section breakdown, confidence indicators, and flags wherever required data was missing rather than invented. Seven editing modes handle the revision pass — simplify, formalize, shorten and the rest — with version comparison so changes are visible, and templates so recurring instruments start from your firm's language, not a generic one.
Draft mode is a guarantee, not a setting
Palmer operates in strict draft mode: it never finalizes, never sends, never files. Fact binding attaches sources to the drafted statements so a reviewer can see where each fact came from; missing data is flagged instead of guessed; approval enforcement means nothing leaves draft status without attorney sign-off. The drafting agent can prepare a letter from retainer to final brief — and cannot deliver any of it.
Court and agency forms, pre-filled
Form filling is where re-keying dies. Portia maps case data already on file to official form fields using a ten-level matching strategy, returning structured output — field, value, source, confidence — so reviewers see where every value came from. Coverage includes 90+ form types: 11 USCIS forms, 5 IRS forms, 73 state court forms and 5 copyright/USPTO forms, on fillable PDFs and scanned overlays alike.
Cross-field validation and consistency checks run after population: fields that depend on each other are verified against each other, and stale-field alerts catch values that have since changed upstream. A worked example from the product: an I-485 filled across 47 fields at 92% average confidence — every field reviewable, every source visible.
The review pipeline
The sequence is deliberate: case data maps to form structure, fields are populated (deterministically where the mapping is certain, with AI assistance where judgment is needed), validation runs, and the form lands in review. Your team inspects each field and its source, edits, overrides or approves individual values, and submits on your terms. Enter information once, and it flows everywhere the firm needs it.
From approved draft to signed document
Once an attorney approves a draft, the remaining path is ordinary practice mechanics: key terms and obligations extracted, version history preserved, e-signatures captured in-browser with an audit trail attached. The AI's job ended at approval; everything after is executed under the same permissions and logging as any other firm record.
Frequently asked questions
Does the AI file anything with a court or agency?
No. FernLex is not a filing service. Every draft and every form goes through your team before it goes anywhere else — nothing is finalized, sent or filed without explicit human approval.
How does form filling know which value goes in which field?
A ten-level matching strategy maps case data to official form fields, and every populated field returns a field-value-source-confidence record. Your reviewers see the source of each value and can edit, override or approve it individually.
What if the AI doesn't know a required value?
It's flagged rather than filled with a guess. Missing-data flagging in drafting and confidence scoring in forms both exist to route uncertainty to a human, not to paper over it.
Can we use our own templates?
Yes. Template management lets recurring documents start from your firm's language, with version comparison tracking edits after generation.
Which plan includes drafting and form automation?
Document and court forms automation start on Professional; the full AI drafting workforce — Palmer and Portia alongside Atlas and Vera — is part of the Business plan.